Fix Piper-related issues caused by CLI interface changes in v1.3.0

[myeungdev]

First of all, thank you for this fantastic project!

I recently discovered this project and tried to get it running, but encountered a few issues preventing it from working correctly. Specifically, when attempting to use TTS with openwebui, all audio output said "Download DIR voices." Additionally, voice downloading was broken, requiring me to manually download the voices and place them in the voices folder.

After investigating, I found that piper has been updated to version 1.3.0, and some of the CLI arguments used in this project are now invalid. The changes in this pull request seem to resolve these issues, and the project is now working as expected.

Here's the reference I used to update the CLI arguments:
https://github.com/OHF-Voice/piper1-gpl/blob/main/docs/CLI.md

I understand this project is no longer actively maintained, but I'm submitting this information in hopes it will be helpful to others who encounter the same problems.

For those running into these issues, you can alternatively clone my fork and build the Docker image:
https://github.com/myeungdev/openedai-speech/tree/piper-v1.3.0

[matatonic]

Thank you! I know a number of people still use it so will update it with this patch.
I see you removed the --update-voices calls for piper, is that no longer required? I called it to update the list of voices before downloading any specific voices.

[myeungdev]

To be honest, I wasn't entirely sure what the --update-voices calls did and could find very little documentation about this argument. All I know is that it downloads the latest voices.json during startup. In piper v1.3.0, this argument is no longer mentioned (doc), and the voice download command seems to have changed entirely, so I assume that --update-voices is no longer needed.

Is there a way to verify this change? So far, I've only verified it by running bash download_voices_tts-1.sh en_US-ryan-high, which worked as expected, leading me to believe the change is correct.

[matatonic]

It's been a while since I used this to be honest and I can't recall exactly why it's there... but I think it was due to the version of voices.json being old when installed and needing to update it before downloading the listed voices, otherwise it just fails the first time. Maybe it's fixed, so not needed.

I do still want to test the build for any breakage, it's going to push new artifacts for more than just piper users, and I know there have been some problems for some people with xtts over the last months so it may take me a minute (*a while) before I release a new version.

Thanks again.